Welcome to Unyo
A 60-second orientation: what Unyo is, the 10 agents you'll be working with, the shared memory behind them, and the three things to do first.
Questa pagina non è ancora disponibile nella tua lingua - stiamo mostrando la versione in inglese.
Unyo is a team of ten AI coworkers for your business. You don't configure them, you talk to them - and they come back with real, editable work: a post draft with its image, a reply sitting in your inbox waiting for one click, a document with citations, a spreadsheet, a live website.
Two things make that work, and they're worth sixty seconds of your time.
1. Ten agents, not one chatbot
Each agent owns one job, and owns it exclusively. This isn't ten personalities on top of the same assistant - each one has a different toolbox, enforced at the runtime level. Ashley genuinely cannot open your inbox. Only Maya can.
| Agent | Owns |
|---|---|
| Maya | Email and calendar - the only agent with your inbox |
| Ashley | Social media - drafts, scheduling, publishing, analytics |
| Lucy | Content writing - research-then-write, with real citations |
| Tyron | Images and audiovisual |
| Blake | Data analysis - runs real Python, not estimates |
| Alex | Sales |
| Riley | Product |
| Sam | Customer support |
| Ema | Personal development, and your "what's on my plate" snapshot |
| Steve | Websites - describe one, watch it get written, publish it |
Ask an agent for something outside its job and it hands the task to the right coworker. The coworker opens its own conversation and starts work in the background - you get a notification the moment it takes the task, and the conversation fills in as it goes.
Agents draft. You decide.
No agent sends an email or publishes a post on its own - there is no auto-send setting to find, because unattended sending doesn't exist in Unyo, by construction. Steve only puts a site online when you explicitly ask him to publish it. Everything arrives as something you read, edit and approve. That's the boundary, and it's deliberate.
2. One memory, shared by all ten
Every agent reads the same Neural Core - your brand, products, contacts, files and knowledge, plus what you've told them along the way. It's not a per-chat memory that forgets when you close the tab, and it's not ten separate memories that drift apart.
Tell Maya "we never use emojis" and Ashley knows on her next run. Tell Ema your company name once, and Riley, Lucy and Blake all have it. Agents also share a view of your day: what happened across all your conversations in the last 18 hours.
A few properties worth knowing:
- Using your memory is free. Every Neural Core tool costs zero credits.
- Deleted stays deleted. Remove something and the system is explicitly told not to re-learn it.
- It won't learn the AI's own inventions. Facts are extracted from what you say, never from what an agent said back to you.
- It can't be forged. Only the extraction service can write to your long-term memory - nothing reaching your account from the outside can put words in it.
Read more: The Neural Core.
What to do first
Three steps, roughly ten minutes.
- Fill in your Neural Core. Open the Neural Core from the sidebar and add your company description, your brand voice, and a product or two. Everything you put here, all ten agents can use. This is the single highest-leverage ten minutes in the product - an agent with no Neural Core writes generic copy, and an agent with one doesn't.
- Connect one integration. Open Integrations from the sidebar and connect the one account you'd actually use today - Gmail or Outlook for Maya, Instagram or LinkedIn for Ashley. Don't connect all sixteen. Connect one, get one result, then add more.
- Ask for one real thing. Open a chat with the agent that owns it and ask for something you were going to do anyway.
Draft an Instagram post announcing that we're open on Sundays from now on. Match our brand voice and include an image.
You'll get an editable draft card - caption, hashtags and a generated image - not a wall of chat text. Change it, then publish when you're happy.
Watch it work
While an agent runs, it declares its plan and ticks each step off live. Expand any finished step to read what it actually found or wrote. Close the tab mid-run and the work keeps going server-side - come back and it's waiting for you.
What it costs to try
You start with 2,500 free credits and no card, and you're billed on the real cost of the work, not per message.
Reading, searching, drafting and planning are free - 73 of the 105 priced tools cost zero credits. You pay when something expensive actually happens, like generating an image or publishing a website. And if a paid action fails, the credits come back automatically.
Read more: Credits and billing · Plans and top-ups.
Where things live
| You want to… | Go to |
|---|---|
| Talk to an agent | Chat, from the sidebar or the home input bar |
| Add brand, products, contacts, files | Neural Core, from the sidebar |
| Connect Gmail, Instagram, Slack… | Integrations, from the sidebar |
| See what's scheduled and what already ran | Automations, from the sidebar |
| Change language, plan, credits | Settings - it opens as a modal, from the sidebar |
| Launch a one-click workflow | Skills, on the agents that have them |
Settings is a modal
There's no settings page to navigate to - the gear in the sidebar opens everything in place: Account, Preferences, Automations, Subscription and Credits.
Languages
Set them independently:
- 6 interface languages - English, French, Spanish, German, Italian, Turkish.
- 11 agent reply languages - your agents answer in the language you choose, no matter what language you type in. You can also just ask mid-conversation to switch.
- 11 voice languages.
Changing your agent language changes what the agents say. It never changes your interface.
FAQ
Which agent should I start with?
The one who owns the job you're avoiding. If it's your inbox, that's Maya. If it's posting consistently, Ashley. If it's writing something long, Lucy. If you're not sure, ask any of them - if it's not their job, they'll hand it to the coworker whose job it is.
Will an agent send an email or publish a post without me?
No, and there's no setting that would let it. Sending and publishing require you. Even a scheduled, recurring task prepares a draft and stops there - it never sends on its own. If you ever see copy anywhere promising unattended sending, it's wrong.
Do I have to explain my business to each agent separately?
No. That's what the Neural Core is for. Tell it once - brand, products, tone, contacts - and all ten agents read the same thing. What you tell one agent in conversation also becomes memory the others can use.
Can I ask any agent to build me a website?
No - websites are Steve's, and Steve is reached directly. Open a conversation with Steve and describe what you want. Other agents can hand work to each other, but a website isn't something they can pass along; that one you start yourself.
What happens if an agent needs an account I haven't connected?
It doesn't fail - it pauses and shows you a Connect card. Connect the account and the run picks up exactly where it stopped, with no need to retype your request. If you leave it more than a day, the run won't auto-resume; just ask again.
Is my data used to train AI models?
No. Every model Unyo uses runs under a no-training guarantee. Your access tokens are encrypted at rest, every table in the database is protected by row-level access rules, and deleting your account purges your data rather than archiving it. See Security and privacy.
Troubleshooting
"The agent said it can't do that." It's probably not its job. Toolboxes are enforced, not suggested - Ashley can't read your inbox even if you ask nicely. Ask her to hand it to Maya, or open Maya directly.
"It answered but didn't use anything about my business." Your Neural Core is likely empty. Add your company description and brand voice, then ask again - the difference is immediate.
"An agent asked me to connect something I already connected." Some agents only get certain capabilities in certain modes, and one connection per provider is what's stored - you can have Gmail and Outlook, but not two Gmail accounts. Check the Integrations page to see what's actually linked.
"The run stopped part-way through a long task." Interactive runs have a turn budget. Ask for the remaining piece as a follow-up message in the same conversation - the agent still has the context.
Where to go next
- Getting started - the longer setup walkthrough.
- The AI agents - what each of the ten owns.
- How agents work - steps, transparency, what happens when you close the tab.
- The Neural Core - the shared memory in detail.
- Credits and billing - what's free and what isn't.